string KmerAffectAnalyser::toString() const{
string kmer;
for (size_t i = 0; i < affectations.size(); i++) {
kmer += affectations[i].toString();
#ifdef DEBUG_KMERS
kmer += ": "+spaced(seq.substr(i,kms.getS()), kms.getSeed())+"\n";
#endif
}
return kmer;
}
void FuzzySearchImpl::query(const QString &req, QVector<Service::Item *> *res) const
{
QVector<QString> words;
for (QString &word : req.split(QRegExp("\\W+"), QString::SkipEmptyParts))
words.append(word.toLower());
QVector<QMap<Service::Item *, unsigned int>> resultsPerWord;
// Quit if there are no words in query
if (words.empty())
return;
// Split the query into words
for (QString &word : words)
{
unsigned int delta = word.size()/3;
// Get qGrams with counts of this word
QMap<QString, unsigned int> qGrams;
QString spaced(_q-1,' ');
spaced.append(word.toLower());
for (unsigned int i = 0 ; i < static_cast<unsigned int>(word.size()); ++i)
++qGrams[spaced.mid(i,_q)];
// Get the words referenced by each qGram an increment their
// reference counter
QMap<QString, unsigned int> wordMatches;
// Iterate over the set of qgrams in the word
for (QMap<QString, unsigned int>::const_iterator it = qGrams.cbegin(); it != qGrams.end(); ++it)
{
// Iterate over the set of words referenced by this qGram
for (QMap<QString, unsigned int>::const_iterator wit = _qGramIndex[it.key()].begin(); wit != _qGramIndex[it.key()].cend(); ++wit)
{
// CRUCIAL: The match can contain only the commom amount of qGrams
wordMatches[wit.key()] += (it.value() < wit.value()) ? it.value() : wit.value();
}
}
// Allocate a new set
resultsPerWord.push_back(QMap<Service::Item *, unsigned int>());
QMap<Service::Item *, unsigned int>& resultsRef = resultsPerWord.back();
// Unite the items referenced by the words accumulating their #matches
for (QMap<QString, unsigned int>::const_iterator wm = wordMatches.begin(); wm != wordMatches.cend(); ++wm)
{
// // Do some kind of (cheap) preselection by mathematical bound
// if (wm.value() < qGrams.size()-delta*_q)
// continue;
// Now check the (expensive) prefix edit distance
if (!checkPrefixEditDistance(word, wm.key(), delta))
continue;
for(Service::Item * item: _invertedIndex[wm.key()])
{
resultsRef[item] += wm.value();
}
}
}
// Intersect the set of items references by the (referenced) words
// This assusmes that there is at least one word (the query would not have
// been started elsewise)
QVector<QPair<Service::Item *, unsigned int>> finalResult;
if (resultsPerWord.size() > 1)
{
// Get the smallest list for intersection (performance)
unsigned int smallest=0;
for (unsigned int i = 1; i < static_cast<unsigned int>(resultsPerWord.size()); ++i)
if (resultsPerWord[i].size() < resultsPerWord[smallest].size())
smallest = i;
bool allResultsContainEntry;
for (QMap<Service::Item *, unsigned int>::const_iterator r = resultsPerWord[smallest].begin(); r != resultsPerWord[smallest].cend(); ++r)
{
// Check if all results contain this entry
allResultsContainEntry=true;
unsigned int accMatches = resultsPerWord[smallest][r.key()];
for (unsigned int i = 0; i < static_cast<unsigned int>(resultsPerWord.size()); ++i)
{
// Ignore itself
if (i==smallest)
continue;
// If it is in: check next relutlist
if (resultsPerWord[i].contains(r.key()))
{
// Accumulate matches
accMatches += resultsPerWord[i][r.key()];
continue;
}
allResultsContainEntry = false;
break;
}
// If this is not common, check the next entry
if (!allResultsContainEntry)
continue;
// Finally this match is common an can be put into the results
finalResult.append(QPair<Service::Item *, unsigned int>(r.key(), accMatches));
}
}
else // Else do it without intersction
{
for (QMap<Service::Item *, unsigned int>::const_iterator r = resultsPerWord[0].begin(); r != resultsPerWord[0].cend(); ++r)
finalResult.append(QPair<Service::Item *, unsigned int>(r.key(), r.value()));
}
// Sort em by relevance
std::sort(finalResult.begin(), finalResult.end(),
[&](QPair<Service::Item *, unsigned int> x, QPair<Service::Item *, unsigned int> y)
{return x.second > y.second;});
for (QPair<Service::Item *, unsigned int> pair : finalResult){
res->append(pair.first);
}
}
